Exclusive Content   Join Now

Familiarity breeds ‘dogs’

Oct 19, 2004 12:49 AM

If you are trying to predict how an upcoming game will play out, an obvious piece of data to consider is how the two teams have done when they’ve played each other in the past.

The question though is whether history matters in the NFL, given that turnovers are a constant of pro football life, with new players, coaches, schemes, and even rules coming into the equation.

To ravage a common saying, "Are handicappers who are ignorant of head-to-head history doomed to miss out when it repeats?" Perhaps our investigation may find that there is no predictive value to past meetings, in which case it would be one less thing to worry about!

There are reasons to believe that one team might have a long term advantage over another. Certainly division rivals that play twice a year are more familiar with each other’s personnel at a higher level than with teams they may only play every few years.

Perhaps coaching becomes more important in such games, though a counter argument could be made that a coach’s influence could be strongest when the two teams know little about each other. Past results may paint an extra layer of motivation into the rematches which could possibly have pronounced effects. Still, we don’t dabble too much with rationalizations. Let’s look at the actual numbers!

A simple first pass look, requiring teams to have met at least 10 prior times since 1990 shows that there isn’t much value in knowing straight-up head-to-head.

A couple of subsets do show promise:

”¡ Home dogs with a 70 percent record in prior meetings are 25-13

”¡ Underdogs with a 70 percent record getting 3½ points or more are 51-26

Another way to look at head-to-head results though is to focus on spread performance instead of outright wins. In fact, it’s slightly profitable to have played every single matchup where one team had covered 60 percent or more of the past meetings. Home dogs (50-33) and big dogs (93-60 getting 3½+) make the most headway. All underdogs are 159-119 (57.1%).

Of course, in this analysis we have lumped all games between two teams together. However, a fairly common theory on head-to-head handicapping is that you should look at home/away specific results for the real story. Taking that step produces the following tables (minimum 5 prior meetings at the home team’s location).

Of the four looks we’ve researched, the best has been using spread history all for both home/away together. The splits for the specific location were not an improvement!

An alternative way of looking at the history would be to use ”˜net points’ instead of a won/lost outcome. About the only subset of note is home teams that have on average beat the spread by 6+ points against a particular opponent are 103-72 ATS (59 percent).

In summary, you are probably not missing much if your handicapping doesn’t include past head-to-head performance as a factor of your analysis. There is some evidence that good prior spread performance against a team over the last 10 meetings is a positive, especially a team with success as an underdog.

There are a couple of other areas of interest for head-to-head history — teams meeting for the second time in a year, and "over / under" performance.

We will tackle these next week.